缺陷检测模型训练方法、缺陷检测方法及装置
By using a defect detection model training method with cascaded attention mechanisms, the problems of insufficient detection accuracy and difficulty in data collection during high-altitude inspections are solved, achieving efficient defect identification and real-time analysis, and improving detection accuracy and background recognition capabilities.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- LANZHOU UNIV
- Filing Date
- 2025-11-07
- Publication Date
- 2026-07-17
AI Technical Summary
Existing high-altitude inspection defect detection methods suffer from insufficient detection accuracy, lack of real-time performance, and difficulty in data collection, especially in complex environments where efficient defect identification and real-time analysis are difficult to achieve.
A defect detection model training method using a cascaded attention mechanism is adopted. Through the backbone network, channel attention branch and spatial attention branch, high-quality defect prediction boxes and background false detection boxes are generated. The target localization error is dynamically adjusted to enhance the learning ability of the background region. The model is trained using a small number of multimodal defect sample images.
It improves detection accuracy, reduces false detection and false negative rates, enhances background recognition accuracy, and enables real-time and accurate defect detection of multimodal image data collected by high-altitude UAVs.
Smart Images
Figure CN121392671B_ABST
Abstract
Citation Information
Patent Citations
Unmanned aerial vehicle defect detection method based on channel space fusion cross attention
CN119323740A
Circuit board defect detection method and equipment based on improved YOLOv8
CN120235837A